Go Small to Go Big: Small Wins Prove AI Isn’t a Bubble

Go Small to Go Big: Small Wins Prove AI Isn’t a Bubble

The overarching narrative of AI being a bubble is rooted in the idea that behind all the NVIDIA GPUs shipping to hyperscalers there is a nearly non-existent demand for AI.

At first glance, the idea that Meta Platforms Inc., Alphabet Inc.’s Google, Amazon.com Inc., Tesla Inc., and other companies would spend more than $100 billion in capex on AI infrastructure (sell-In) without a clear view of what customers are going to do with it (sell-out) is almost absurd. However, there are very smart people in Venture Capital and financial services at firms like Sequoia and Goldman Sachs that are suggesting AI, at least at this current juncture, is a bubble.

Is there an AI Bubble?

The answer depends on the timetable by which you measure AI and its impact on industries.

First, historically speaking, AI algorithms are more than four decades old, and the tech industry has been investing heavily in variants of artificial intelligence such as machine learning, deep learning, neural networks, and more extensively for at least 10 to 15 years.

Generative AI, and its late 2022 arrival, changed the timeline and gave a glimpse of what felt like a more prescient type of AI that generated text, images, then video, and now a variety of outputs — providing a sneak peek at what we refer to as Artificial General Intelligence, the holy grail or end of days depending on whom you talk to.

Today, the market is seeking to better understand how the sell-in of GPUs and XPUs at a pace of more than $100 billion in 2024, and growing, is going to convert to value outside of the tech industry.

For instance, AI is showing up at banks, hospitals, hotels, manufacturing plants, transportation, retail, and more. The bear thesis is AI is mostly hype, with limited evolution from the data optimization tactics of the past 15 years and that customers aren't going to pay more for SaaS services with Generative AI capabilities or new PCs and smartphones with AI features.

Our intelligence data suggests otherwise. What many aren’t accounting for in the AI calculus is the timeline that AI will be digested and deployed, and how that will vary across industries and lead to a variance in time-to-value for many business use cases.

Our data is also showing triple-digit percent increases in multimillion-dollar proof of concepts (POCs) in 2024 and mid- to high- double-digit percentage growth of AI use in key industries like financial services, healthcare, and telco growing near 40% CAGR over the next five years. Use cases like object detection and conversational AI are also seeing more than 35% growth (five-year CAGR), which means the silicon investment and infrastructure build out will have to convert to software and industry use cases. However, it is more of a drip right now. And once the dam breaks, it will become a waterfall.

I've regularly referenced the three AI network effects, and I’ll refer back to them here once again. We are seeing most of the AI impact on the economy in the n0 and n1 effects with increasing small wins for the n2 node. All of this provides a big opportunity for the System Integrator community that is building these POCs and supporting the buildout of AI within the enterprise chip design and manufacturing network (NVIDIA, AMD Inc., Synopsys Inc., TSMC, Arm Holdings, Intel Corp., etc.)

Network One Infrastructure buildout (Dell, Lenovo, Super Micro Computer Inc., AWS, Microsoft Azure, Google Cloud, Oracle Corp., etc.)

Network Two Enterprise Software and SaaS + Devices at the edge (Salesforce Inc., Microsoft Corp. apps, Palantir Technologies Inc., SAP, Oracle, ServiceNow Inc., IBM Corp., HP Inc., Dell Inc., Samsung Electronics, Apple Inc., etc.)

Network Three Industry Adoption and Implementation (financial services, manufacturing, healthcare, etc.)

In conclusion, AI and mega cap tech have a different horizon for AI than most investors and certainly traders.

Businesses have been extracting value in AI for years, and generative AI has accelerated demand, increased data volumes, and increased pressure on enterprises to implement AI faster to stay competitive. But the expense of build out and the complexity of the data estate are substantial challenges for most companies, and the tech industry does have an important point to make that AI features are incremental. Customers and consumers are willing to pay for AI features, especially those that have free (often open source) alternatives.

There isn’t really a bubble around AI, but there also aren’t enough big wins for analysts and investors to fully see the economic impact beyond the data center.

I see that changing in the next 18 months. Some may think that is too long, so we’ll have to settle for small wins that keep on coming or perhaps another way of looking at it is when it comes to AI implementations it will be slow at first, and then (nearly) all at once.

Article first seen on Forbes : Go Small To Go Big: Small Wins Prove AI Isn’t A Bubble (forbes.com)

I believe it's very real. Interesting article. . Fyi Sandy Carter

Like
Reply
Mark Donnigan

Marketing Leader and Tech Company Builder

3mo

The significant investments in AI by tech giants mirror the strategic foresight highlighted by Clayton Christensen. Much like how disruptive innovations initially encounter skepticism, AI's potential is vast yet gradual in adoption. Financial services and healthcare sectors already show impressive growth, indicating that these advances may well pave the way for widespread transformation across industries.

Ben Parisi

Founder, Benten Alignment Inc. | Democratizing AI Systems with Web3 | Aligning Humanity

3mo

Thought-provoking post, Daniel. While tech giants are making huge investments, I'm curious about your thoughts on the role of decentralization in the AI landscape. I have a feeling that community-driven, open-source AI development will provide a counterbalance to big tech dominance and will start to accelerate adoption more and more over the next 18 months, potentially democratizing AI capabilities and fostering innovation from a broader base of contributors.

David Nicholson

The Futurum Group/ Six Five Media

3mo

The part of the equation that makes this feel bubbly to me is the absolute consensus that technology in the form of AI will deliver the economy from current monetary and energy policy induced woes. It’s axiomatic that the production possibilities curve can be pushed outward by tech advances that improve efficiency. Can AI be a big enough sponge to soak up the current spill? I am cautiously optimistic.

Larry Livingston

Founder@BizLeague / The Business Network: Reimagined

3mo

Daniel Newman I don't believe it's a bubble. It's here to stay, like that distant cousin sleeping on your couch. He said it was temporary, but 6 months later, he's still there and talking about his next big idea. 😎

Like
Reply

To view or add a comment, sign in

Insights from the community

Explore topics